Search

In the presented paper a four level control concept for gas networks will be introduced. The first level deals with the compressor control and is completely local. It consists of a set-point control for the mass flow and a security control to prevent surging. The local operating optimization and a global network observer are placed in the second leve14 The former determins how the compression stages in a station have to be connected and the flow rate of each of them. From the latter the unknown states in the net are obtained. These are necessary for a look ahead simulation. Furthermore the observer can be used for a leak detection . Using the short range optimization in the third level the optimal compression ratios and flow rates are obtained taking the dynamics of the network in to consideration. The strategic planning for the gasnet operation is carried out in the fourth level. Here normally a steady state model is required.

A test apparatus has been devised for testing of wet fibre-reinforced composites at elevated temperatures. The apparatus allows the determination of the compression strength of a laminate at temperatures greater than 100°C without drying out of the specimens. Test results for two different laminates; with bismaleimide and epoxy resin matrices, showed that prevention of drying out has a marked effect on the compression strength measured for ‘wet’ specimens

The instationary gasflow in large gas transmission networks is modelled by a coupled system of partial differential equations. The simulation model for a single pipeline can be normalized in a way that exploits the dominant role of the mach number (M) and the tube factor (RW) for characterizing the flow's and the pipeline's parameters. Introducing a spatial discretization leads to a nonlinear system of ordinary differential equations that - dependent on the values of M and RW - is stiff. In the first stage the discretized model for a single pipeline is reduced by the means of ‘Singular Perturbation’. Inspecting the eigenvalues of the appertaining Jacobian Matrix shows a good representation of the slow modes. The network's dynamic model consists of several reduced models for a single pipeline and an auxilary flow and pressure condition at the nodes. By defining time constants it is shown that also the network model may be stiff. In the second stage the network's dynamic model is reduced by explicit and implicit ‘Singular Perturbation’ methods. Using the reduced model makes it possible to adapt simple explicit integration algorithms which decrease the computing time for simulation.